Predictive Maintenance for Electric Grid

Challenge

One of the largest electric utilities in the United States provides electricity to more than seven million commercial and residential customers in six states, operating an extensive network of more than 36,000 miles of transmission lines and substations.

Prior to engaging C3 AI, production sites relied on time-based maintenance and consequently, experienced unplanned grid downtime and increased costs from emergency responses and service interruptions. The utility saw a need for a new strategy to monitor fleet health, reduce risk, and optimize asset utilization by combining rules-based analytics with machine learning insights.

The utility chose C3 AI Reliability, C3 AI’s market-leading predictive maintenance application, to enable proactive asset lifecycle management and predict asset failures in advance.

Approach

Over 15 months, the C3 AI team partnered with subject matter experts from utility company to unify 10 years of historical and live data from 12 disparate data sources, apply machine learning predict asset failures, and configure the C3 AI Reliability application to monitor 10,000 transformers and 22,000 circuit breakers across 4 operating regions.

With C3 AI Reliability, the utility company have reduced transformer failures by 48%, achieved an estimated $800,000 in annual savings in operations and maintenance costs and estimates over $40M in annual economic value from optimized operational and capital expenditure. Moreover, the utility can shift from scheduled and reactive maintenance to predictive and proactive model, allowing the utility to surface at-risk assets in advance and optimize replacement schedules and costs.

With the success of initial deployment, the company has also adopted C3 AI Ex Machina, a drag-and-drop data analysis and ML development tool, to leverage the unified data for ad-hoc analysis and solution prototyping.

About the Company

• $20B+ annual revenue
• 7M+ retail customers
• Active in 6 states

Project Objectives

• Convert from time-based to condition-based maintenance of transformers and circuit breakers
• Apply machine learning to predict asset failures in advance
• Configure C3 AI Reliability user interface to surface AI insights and unified analytics
• Integrate and unify data from 12 disparate data sources (e.g., sensor data, work orders, inspection data, monitoring events)

Project Highlights

• 10 years of historical times series data from 12 sources integrated
• 160 analytics configured to track asset health, criticality, and KPIs for each asset
• 10,000 transformers and 22,000 circuit breakers in-scope
• Deployed to initial operating region from project kick-off within 15 months